SVT-AV1 and libaom settings map
libaom, through its program aomenc, and SVT-AV1 are two encoders for AV1. Their options do the same jobs as x264's and x265's, under other names, on other scales and from other defaults. This page takes eleven controls and shows how each is written in the four encoders, so that a setting understood in one can be found in the others.
Translate a control
Choose what you want to control. The eleven cover speed, rate control, the keyframe interval, lookahead, adaptive quantization, the in-loop filters, grain and tuning. Each row is one encoder's option for it, its starting value at the version linked, and how the value reads. The libaom column uses its library's good-quality defaults: aomenc's usage mode, its pass count, the build and other options can change what takes effect, and for AV1 aomenc runs two passes unless told otherwise, where the library's configuration starts with one. The names differ less than the scales: a quality target of 23 is a different request in every one of the four.
| Encoder | Option, default and how it reads |
|---|
- The four quality controls start at 23, 28, 10 and 35, and libaom's 10 is only stored: its library starts in a bitrate mode, and
--end-usagehas to select a quality mode before the level is used. - Names are shared more than values. Six of the eleven controls are written in only two distinct forms across the four encoders, and none shows fewer than three distinct default labels.
- libaom is the odd one out on modes. Its library starts in variable bitrate at 256 kbit/s with adaptive quantization mode 0; the other three start in constant quality.
- Both AV1 encoders start with a tuning named for peak signal-to-noise ratio, where x264 and x265 start with their psychovisual tools on.
What the AV1 encoders add
- Switches for filters that have no H.264 counterpart. The directional enhancement filter and loop restoration can each be turned off.
- Film grain synthesis as a setting. Both can model grain and signal parameters for the decoder to synthesize from; x264 and x265 have nothing of the kind to switch on. Denoising the coded pictures is a separate switch:
--film-grain-denoisein SVT-AV1, off by default, and--enable-dnl-denoisingin aomenc, on by default in its help text. aomenc's denoising options exist only in builds with denoising, and its--film-grain-tabletakes a file, not a number. - A second adaptive quantization control in libaom.
--deltaq-modeis on by default while--aq-modeis off, and its help text ties mode 1 to another option,--enable-tpl-model. - Usage modes. aomenc takes
--good,--rtor--allintra, and its help gives--cpu-useda different range in each.
What does not translate
- A preset is not a level of quality. SVT-AV1's 8 and x264's medium are both defaults; nothing read here says they spend comparable effort.
- A default mode is not a recommendation. Three of the four start in constant quality and libaom starts at 256 kbit/s; a comparison has to set the mode in all of them.
- The starting objectives differ. x264 and x265 start with psychovisual tools on and the AV1 encoders with a tuning named for a metric; what that tuning changes inside them was not read. Set each encoder's tuning to suit the evaluation you intend, one of the test design pitfalls.
Sources
Read 2026-11-15, at the commits linked.
- libaom at commit 1b7f9ef (GitHub mirror): the option definitions in
av1/arg_defs.cfor--cpu-used,--end-usage,--cq-level,--target-bitrate,--buf-sz,--kf-max-dist,--lag-in-frames,--aq-mode,--deltaq-mode,--enable-cdef,--enable-restoration,--denoise-noise-level,--film-grain-table,--tune,--good; the library defaults inav1/av1_cx_iface.cfor speed, tuning, quality level, the two filters, the two quantization modes, denoising, lag, rate control mode, bitrate, buffer, keyframe distance. Read 2026-11-16: inapps/aomenc.c, the configuration aomenc asks the library for and two passes as its AV1 default;--enable-dnl-denoisingand the build condition on the denoising options. - SVT-AV1 at commit d0d5c7e (GitHub mirror): the option names in
Source/App/app_config.c(--preset,--crf,--rc,--tbr,--buf-sz,--keyint,--lookahead,--aq-mode,--enable-dlf,--enable-cdef,--enable-restoration,--film-grain,--tune); the defaults inSource/Lib/Globals/enc_settings.cfor rate control mode, lookahead, adaptive quantization, preset, keyframe interval, the three filters, film grain, tuning; ranges and meanings fromDocs/Parameters.md: preset, tune, rate control, quality target, bitrate, adaptive quantization, buffer, keyframes, lookahead, film grain and its denoising switch, with that switch's default (read 2026-11-16). - x264 and x265: the sources of their maps, and x265's tunings.
- Not checked: any behavior of SVT-AV1, of which only names, defaults and documentation were read; what the aomenc program changes from the library's defaults, beyond its pass count; the usage modes other than good quality; what any tuning changes in the AV1 encoders; what each adaptive quantization mode does in the AV1 encoders; x265's grain tuning; whether any two presets or quality targets are comparable.